Nuclear power plant diagnostics using artificial neural networks

Enhanced safety, reliability and operability of nuclear power plants may be achieved by the application of neural networks as a diagnostic tool to define the state of the plant at any given time. The paper presents a new neural network methodology, based on the backpropagation learning algorithm, for malfunctions management in nuclear power plants. It is shown that neural networks can be used for identifying the nonlinear dynamic behaviour of nuclear power plant components, and for isolating the origin and extent of a failure, when occurring, using consecutive samplings of sensors readings. (author). 21 refs, 5 figs